]> git.apps.os.sepia.ceph.com Git - ceph.git/commit
msg/async/rdma: use different strategy to reset read/write chunk
authorChangcheng Liu <changcheng.liu@aliyun.com>
Mon, 1 Jul 2019 02:27:45 +0000 (10:27 +0800)
committerChangcheng Liu <changcheng.liu@aliyun.com>
Fri, 23 Aug 2019 03:35:55 +0000 (11:35 +0800)
commit32da5f1d038bb9b8abb8cc7fccd7df3f3e4ceef4
tree0ffe8cc80f96c1b8cb927437848cd46cf33f50fc
parent60a87c9db99fceda9f6d4efee46b9f06175bcec0
msg/async/rdma: use different strategy to reset read/write chunk

When releasing read chunk to pool, the chunk::offset & chunk::bound
should be reset to zero. For write chunk, it's better to reset
chunk::offset to zero and chunk::bound to chunk length which means that
[offset, bound) is writable.

Signed-off-by: Changcheng Liu <changcheng.liu@aliyun.com>
src/msg/async/rdma/Infiniband.cc
src/msg/async/rdma/Infiniband.h
src/msg/async/rdma/RDMAConnectedSocketImpl.cc